Intelligent Bioelectrochemical Systems Enhanced by Machine Learning

PhD Research Project – University of São Paulo (USP)

OVERVIEW: This doctoral research investigates the integration of Bioelectrochemical Systems (BES) with machine learning methodologies to enhance predictive performance, system optimization, and electrochemical understanding. The project focuses on microbial fuel cells (MFCs) and microbial electrolysis cells (MECs), exploring electron transfer mechanisms, catalytic behavior, and performance scalability. By combining experimental electrochemistry with data-driven modeling, the research aims to establish hybrid predictive frameworks capable of describing and optimizing bioelectrochemical processes under varying operational conditions.

Research Objectives

• Analyze electrochemical performance of MFCs and MECs

• Model current density, voltage behavior, and internal resistance

• Develop predictive models using machine learning algorithms

• Apply explainable AI techniques to interpret electrochemical patterns

• Evaluate sustainability and energy efficiency metrics

Methodology

• Experimental setup of microbial electrochemical systems

• Electrochemical characterization (CV, polarization curves, power density analysis)

• Real-time data acquisition systems

• Machine learning modeling (regression and predictive frameworks)

• Model interpretability using feature importance analysis

Innovation

This research proposes a hybrid approach that integrates physicochemical modeling with machine learning, bridging experimental electrochemistry and artificial intelligence. The framework enhances predictive reliability and contributes to the development of intelligent, self-optimizing bioelectrochemical platforms.

Impact

• Sustainable energy generation

• Hydrogen production optimization

• Environmental remediation technologies

• AI-driven scientific modeling in electrochemical systems

Computational Linguistic Modeling for Epistemological Analysis of Scientific Discourse

Interdisciplinary Research Project – Linguistics, AI & Philosophy of Science

OVERVIEW: This project applies Natural Language Processing (NLP) techniques to analyze the epistemological evolution of scientific discourse in bioelectrochemical systems research. By constructing and modeling a large corpus of scientific publications, the study investigates thematic shifts, lexical density, semantic structures, and discursive patterns over time. The objective is to understand how scientific knowledge evolves linguistically and conceptually.

Research Objectives

Construct a structured corpus of BES scientific literature
Identify thematic clusters using topic modeling (BERTopic, LDA)
Analyze semantic-pragmatic patterns and hedging strategies
Detect epistemological transitions over a 25-year research period
Quantify discourse evolution using computational metrics

Methodology

Corpus construction and preprocessing
Embedding generation using transformer-based models
Topic modeling and dimensionality reduction (UMAP)
Supervised and unsupervised classification
Statistical analysis of lexical and semantic indicators

Innovation

The project reframes NLP as an epistemological instrument rather than merely a computational tool. By modeling scientific discourse computationally, it reveals structural transformations in knowledge production and research paradigms.

Impact

Advancement of digital epistemology
Computational analysis of scientific paradigms
Integration of linguistics and engineering research
Transferable framework for scientometric intelligence systems

Hybrid Semantic Retrieval and Explainable AI for Latin Forensic Prose

Digital Philology & AI Research Project

OVERVIEW: This project develops a hybrid semantic retrieval architecture for classical Latin forensic prose, integrating sparse lexical models, dense embeddings, and cross-encoder re-ranking systems. The objective is to enhance intertextual detection, semantic alignment, and rhetorical pattern identification while maintaining explainability. The framework bridges digital philology and modern information retrieval systems.

Research Objectives

Develop hybrid retrieval systems (sparse + dense models)
Detect intertextuality and rhetorical parallels in Latin texts
Implement cross-encoder re-ranking for semantic refinement
Apply explainable AI for alignment-based interpretability
Evaluate retrieval performance in historical legal corpora

Methodology

Corpus preparation of Latin forensic texts
Embedding generation using contextual language models
Sparse-dense hybrid retrieval pipelines
Cross-encoder semantic scoring
Alignment visualization for interpretability

Innovation

The project situates Latin forensic prose as a rigorous evaluation environment for hybrid semantic retrieval systems. It contributes both to digital humanities and to transferable architectures applicable in modern legal and institutional AI systems.

Impact

Advancement of digital philology
Explainable AI for historical corpora
Transferable retrieval frameworks for legal AI
Integration of classical studies and computational linguistics